Scenario Analysis and Anomaly Detection of Locational Marginal Price in Electricity Market

LMP is an important price signal in electricity market. However, it can be abnormal due to malicious tempering or unexpected system state. Such anomalies are difficult to artificially identify, which brings difficulty to market regulation and transparency. In this paper, the problem of LMP abnormal detection is formulated and an efficient method based on convolutional neural network (CNN) is proposed to raise alarm on abnormal LMP matrices. Then quantile regression forest (QRF) is used to forecast the electricity prices to identify specific nodes and time intervals in the abnormal matrices. In numerical experiments, comparison with ANN and SVM are conducted. Results show that CNN has higher sensitivity and accuracy for identifying abnormal LMP matrices. This paper also study the smallest changes in LMP that CNN can identify and discusses its feasibility in systems of different scales. Hopefully the research results of this paper can provide useful experience for LMP abnormal detection, improve the level and efficiency of market supervision, and maintain the normal market order.